Device Orientation Independent Human Activity Recognition Model for Patient Monitoring Based on Triaxial Acceleration

Tracking a person’s activities is relevant in variety of contexts, from health and group-specific assessments, such as elderly care, to fitness tracking human–computer interaction. In clinical context, sensor-based activity could help monitor patients’ progress or deterioration during their hospitalization time. However, routine hospital devices face displacements position orientation caused by incorrect device application, physical peculiarities, day-to-day free movement. These aspects can significantly reduce algorithms’ performances. this work, we investigated how shifts impact Human Activity Recognition (HAR) classification. To reach purpose, propose an HAR model based on single three-axis accelerometer that be located anywhere the participant’s trunk, capable recognizing multiple movement patterns, and, thanks data augmentation, deal with displacement. Developed models were trained validated using acceleration measurements acquired fifteen participants, tested twenty-four which twenty different study protocol for external validation. The obtained results highlight changes algorithm potential simple wearable sensor augmentation tackling challenge. When applying small rotations (<20 degrees), error baseline non-augmented steeply increased. On contrary, even when considering ranging 0 180 along frontal axis, our reached f1-score 0.85±0.11 against equal 0.49±0.12.